HAL Id: hal-01960237
https://hal.umontpellier.fr/hal-01960237
Submitted on 19 Dec 2018
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
insights into Plasmodium vivax biology and evolution
Aude Gilabert, Thomas Otto, Gavin Rutledge, Blaise Franzon, Benjamin
Ollomo, Céline Arnathau, Durand Patrick, Nancy Moukodoum, Alain-Prince
Okouga, Barthélémy Ngoubangoye, et al.
To cite this version:
Aude Gilabert, Thomas Otto, Gavin Rutledge, Blaise Franzon, Benjamin Ollomo, et al.. Plasmodium vivax-like genome sequences shed new insights into Plasmodium vivax biology and evolution. PLoS Biology, Public Library of Science, 2018, 16 (8), pp.e2006035. �10.1371/journal.pbio.2006035�. �hal-01960237�
Plasmodium vivax-like genome sequences
shed new insights into Plasmodium vivax
biology and evolution
Aude Gilabert1☯¤, Thomas D. Otto2,3☯, Gavin G. Rutledge2, Blaise Franzon1,
Benjamin Ollomo4, Ce´line Arnathau1, Patrick Durand1, Nancy D. Moukodoum4, Alain-Prince Okouga4, Barthe´le´my Ngoubangoye4, Boris Makanga4, Larson Boundenga4, Christophe Paupy1,4, Franc¸ois Renaud1, Franck Prugnolle1,4*, Virginie Rougeron1,4*
1 MIVEGEC, IRD, CNRS, University of Montpellier, Montpellier, France, 2 Wellcome Trust Sanger Institute,
Wellcome Trust Genome Campus, Cambridge, United Kingdom, 3 Institute of Infection, Immunity and Inflammation, University of Glasgow, College of Medical, Veterinary and Life Sciences, Glasgow, United Kingdom, 4 Centre International de Recherches Me´dicales de Franceville, Franceville, Gabon ☯These authors contributed equally to this work.
¤ Current address: UMR PVBMT, CIRAD, St Pierre, La Re´union, France
*rougeron.virginie@gmail.com,virginie.rougeron@ird.fr(VR);franck.prugnolle@ird.fr(FP)
Abstract
Although Plasmodium vivax is responsible for the majority of malaria infections outside Africa, little is known about its evolution and pathway to humans. Its closest genetic relative, P. vivax-like, was discovered in African great apes and is hypothesized to have given rise to P. vivax in humans. To unravel the evolutionary history and adaptation of P. vivax to differ-ent host environmdiffer-ents, we generated using long- and short-read sequence technologies 2 new P. vivax-like reference genomes and 9 additional P. vivax-like genotypes. Analyses show that the genomes of P. vivax and P. vivax-like are highly similar and colinear within the core regions. Phylogenetic analyses clearly show that P. vivax-like parasites form a geneti-cally distinct clade from P. vivax. Concerning the relative divergence dating, we show that the evolution of P. vivax in humans did not occur at the same time as the other agents of human malaria, thus suggesting that the transfer of Plasmodium parasites to humans hap-pened several times independently over the history of the Homo genus. We further identify several key genes that exhibit signatures of positive selection exclusively in the human P. vivax parasites. Two of these genes have been identified to also be under positive selection in the other main human malaria agent, P. falciparum, thus suggesting their key role in the evolution of the ability of these parasites to infect humans or their anthropophilic vectors. Finally, we demonstrate that some gene families important for red blood cell (RBC) invasion (a key step of the life cycle of these parasites) have undergone lineage-specific evolution in the human parasite (e.g., reticulocyte-binding proteins [RBPs]).
a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS
Citation: Gilabert A, Otto TD, Rutledge GG, Franzon B, Ollomo B, Arnathau C, et al. (2018) Plasmodium
vivax-like genome sequences shed new insights
into Plasmodium vivax biology and evolution. PLoS Biol 16(8): e2006035.https://doi.org/10.1371/ journal.pbio.2006035
Academic Editor: Andrew Read, Pennsylvania State University, United States of America
Received: March 14, 2018 Accepted: August 7, 2018
Published: August 24, 2018
Copyright:© 2018 Gilabert et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability Statement: All sequences have been submitted to the European Nucleotide Archive. The accession numbers of the raw reads are inS1andS2Tables. The genome assemblies of Pvl01 and Pvl06 can be found on the Dryad Repository:https://datadryad.org/resource/doi:10. 5061/dryad.32tm1k4.2. The two genome assemblies and annotations are available at the following ftp address:ftp://ftp.sanger.ac.uk/pub/ pathogens/Plasmodium/P_vivaxLike/. The dataset used to produce the reference allele frequency distributions are available in the following Dryad
Author summary
Among the 5 species responsible for malaria in humans,Plasmodium vivax is the most
prevalent outside Africa. It causes severe and incapacitating clinical symptoms with signif-icant effects on human health. Yet little is known about its evolution, adaptation, and emergence in humans. The recent discovery in African great apes of its closest known rel-ative—P. vivax-like—may help resolve the evolutionary history of P. vivax. This study
aims to characterize the genome ofP. vivax-like isolated from infections of the closest ape
relative to humans to get a better understanding of the evolution of this parasite. A total of 11P. vivax-like samples were obtained from infected chimpanzee blood samples and an
infected mosquito collected in Gabon. We generated, and here present, the first 2 new genomes ofP. vivax-like and 9 additional draft sequences. Genome-wide analyses provide
new insights into the biology and adaptive evolution ofP. vivax to different host species.
Indeed, they highlight lineage-specific evolution of some gene families involved in key steps of the life cycle ofP. vivax. Analyses also revealed that the divergence between P. vivax and P. vivax-like occurred before the one between P. falciparum and its sister species P. praefalciparum.
Plasmodium vivax is responsible for the majority of malaria infections in humans outside
sub-Saharan Africa [1]. Traditionally,P. vivax has been neglected because it causes lower mortality
in comparison withP. falciparum [2,3]. Its ability to produce a dormant liver-stage form (hyp-nozoite), responsible for relapsing infections, makes it a challenging public health issue for malaria elimination. The recent emergence of antimalarial drug resistance [4] as well as the discovery of severe and even fatal human cases [2,5,6] has renewed interest in this enigmatic species, including its evolutionary history and its origin in humans.
Earlier studies placed the origin ofP. vivax in humans in Southeast Asia (“Out of Asia”
hypothesis) based on its phylogenetic position in a clade of parasites infecting Asian monkeys [7]. At that time, the closest known relative ofP. vivax was considered to be P. cynomolgi, an
Asian monkey parasite [8]. However, this hypothesis was recently challenged with the discov-ery of anotherPlasmodium species, genetically closer to P. vivax than P. cynomolgi, circulating
in African great apes (chimpanzees and gorillas) [9,10]. This new lineage (hereafter referred to asP. vivax-like) was considered to have given rise to P. vivax in humans following the transfer
of parasites from African apes [10]. But this “Out of Africa” hypothesis is still debated. More-over, a spillover ofP. vivax-like parasites to humans has been recently documented, thus
mak-ing possible the release of new strains in new host species, specifically in human populations [9].
In this context, it seemed fundamental to characterize the genome of the closest ape relative to the humanP. vivax parasite in order to get a better understanding of the evolution of this
parasite and also to identify the key genetic changes explaining the emergence ofP. vivax in
human populations.
Genome assemblies
ElevenP. vivax-like genotypes were obtained from 2 different kinds of samples: 10 infected
chimpanzee blood samples collected during successive routine sanitary controls of chimpan-zees living in the Park of La Le´ke´di (a sanctuary in Gabon) and 1 infectedAnopheles mosquito
(An. moucheti) collected during an entomological survey carried out in the same park (S1 Table) [11]. For blood samples, white blood cells were depleted using the CF11 method [12] to
repository: doi:10.5061/dryad.32tm1k4. All the other relevant data, alignments, and trees are within the paper and its Supporting Information files.
Funding: Agence Nationale de la Recherche Jeunes Chercheurs Jeunes Chercheuseshttp://www. agence-nationale-recherche.fr/Project-ANR-12-JSV7-0006(grant number ORIGIN Origin, adaptation and evolution of Plasmodium falciparum). Received by FP. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Agence Nationale de la Recherche Tremplin-ERC (TERC3) 2017 http://www.agence-nationale-recherche.fr/Project-ANR-17-ERC3-0002 (grant number EVAD: Evolutionary history and genetic adaptation of Plasmodium vivax). Received by VR. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Laboratoire Mixte Internationalhttps://www.ird.fr/infos-pratiques/ archives/anciens-lmi/lmi-zofac-zoonoses-dans-les- forets-tropicales-humides-d-afrique-centrale- modalites-des-transferts-inter-especes-et-adaptation-des-pathogenes(grant number ZOFAC: zoonoses dans les forêts tropicales humides d’Afrique centrale: modalite´s des transferts inter-espèces et adaptation des pathogènes). Received by FP. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Medical Research Council Doctoral Training Grant (grant number MR/J004111/1). Received by GGR. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Wellcome Trusthttps://www. genomethics.org/overview.html(grant number 098051 Genome Ethics). Received by GGR. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Abbreviations: BLAST, Basic Local Alignment Tool; CI, confidence interval; CIRMF, Centre International de Recherches Me´dicales de Franceville; CLAG, Cytoadherence-linked asexual gene; Cytb,
Plasmodium Cytochrome b; DARC, Duffy Antigen
Receptor for Chemokine; DBP, Duffy-binding protein; ETRAMP, early transcribed membrane protein; GC, guanine–cytosine; G-PhoCS, Generalized Phylogenetic Coalescent Sampler; GTR, general time-reversible; HGAP, Hierarchical Genome Assembly Process; ILS, Incomplete Lineage Sorting; JTT, Jones, Taylor, and Thornton; LRT, likelihood ratio test; MCMC, Markov Chain
reduce the amount of host DNA. After DNA extraction, samples were subjected to whole-genome amplification (WGA) in order to obtain sufficient parasite DNA for library prepara-tion. Sequencing was then performed using short-read Illumina technology. For one sample (Pvl06), long-read sequencing (Pacific Biosciences [PacBio] technology) was performed in order to get a better coverage of regions containing subtelomeric gene families.
Among the 11 samples, 10 presented mixed infections with otherPlasmodium species (S1
andS2Tables). Four samples containingP. gaboni or P. malariae-like co-infections were used
in other studies (seeS1 Table) [13,14]. In order to obtain theP. vivax-like genotypes and to
preclude errors due to co-infections with otherPlasmodium species, sequencing reads were
extracted based on their similarity to the reference genome sequence ofP. vivax, PvP01 [15], and any reads mapping against a reference genome of aLaverania chimpanzee-infecting
spe-cies (e.g., PrG01, PbilcG01, and PGAB01; see Otto and colleagues [13]) were removed (S2 Table). Concerning multiple infections with severalP. vivax-like strains, only 2 (Pvl09 and
Pvl10) seemed to be multiply infected as suggested by the reference allele frequency (RAF) dis-tributions (seeMaterials and methodssection andS1 Fig). In order to avoid any bias in the genomic analysis due to multiple infections with several strains ofP. vivax-like, only 1 variant
was extracted per sample. This was done by considering the allele with the highest frequency in the calling and filtering analysis of Single Nucleotide Variants (SNVs) (seeMaterials and methodssection). Sequencing reads from 2 samples, one obtained using Illumina sequencing, Pvl01, and another using PacBio technology, Pvl06, were used to perform de novo genome assemblies and were annotated to produce reference genomes forP. vivax-like (S1 Table). Of the 2 assemblies, Pvl01 is of considerably higher quality (4,570 one-to-one orthologues to the PvP01 reference genome compared with 2,925 for Pvl06 [Table 1]). Both assemblies consist of 14 supercontigs (corresponding to the 14P. vivax chromosomes)—and 1,176 and
351 unassigned contigs—comprising a total of 27.5 Mb and 18.8 Mb in size, respectively, for Pvl01 and Pvl06, respectively. After annotation with Companion [16], these 2 genomes con-tained 5,532 and 4,953 annotated genes (Table 1). We obtained for the 9 remaining samples between 2.9 and 86 million of reads that paired with theP. vivax PvP01 reference genome
(mean± SD = 2.17 ± 25.71), with a mean depth per high-quality position ranging from 13.98 to 335.1 (mean± SD = 93.43 ± 99.51; seeS3 Table). These genome sequences were used for SNV calling for population genetic and phylogenetic analyses.
Table 1. Genome features of theP. vivax-like Pvl01 (Illumina HiSeq sequenced) and Pvl06 (PacBio sequenced) strains, P. vivax reference strains SalI and PvP01
[15],P. cynomolgi B and M isolates [8,17], andP. knowlesi H strain [18].
P. vivax-like (Pvl01) P. vivax-like (Pvl06) P. vivax (PvP01) P. vivax (SalI) P. cynomolgi (B strain) P. cynomolgi (M strain) P. knowlesi (H strain) Assembly size (Mb) 27.5 18.8 29 26.8 26.2 30.6 24.1 Scaffolds 14 (1,176) 14 (351) 14 14 14 14 14 Overall GC content (%) 44.9 45.8 39.8 42.3 40.3 37.3 37.5 Number of genes 5,532 4,953 6,642 6,690 5,722 6,632 5,188
Gene density (gene/Mb) 201.2 264.5 229 249.6 218.4 216.7 215.3
Coverage#
12.17× 34× - - - -
-One-to-one orthologues with PvP01
4,570 2,925 - 5,178 4,870 5,222 4,804
Unassigned contigs indicated in parentheses.
#Calculated as (L
× N) G, where L is the read length (100 bp were considered for the Illumina sample, Pvl01), N the number of mapped reads, and G the size of the assembly.
Abbreviations: GC, guanine–cytosine; PacBio, Pacific Biosciences.
https://doi.org/10.1371/journal.pbio.2006035.t001 Monte Carlo; PacBio, Pacific Biosciences; PHIST,
Plasmodium helical interspersed subtelomeric;
PIR, Plasmodium interspersed repeat; RAF, reference allele frequency; RBC, red blood cell; RBP, reticulocyte-binding protein; SERA, serine-repeat antigen; SNV, single nucleotide variant; STP1, subtelomeric protein 1; TLS, Total Least Squares; TRAG, tryp-rich antigen; WGA, whole-genome amplification.
Gene synteny and gene composition
Comparing theP. vivax-like reference genomes to those of P. vivax (PvP01 and SalI) [3,15],P. cynomolgi (B and M strains) [8,17], andP. knowlesi (H strain) [18] reveals several similarities, including a similar guanine–cytosine (GC) content and extensive collinearity and conservation of gene content and organization (Table 1). TheP. vivax-like core genome sequences are
completely syntenic to theP. vivax PvP01 reference genome sequence (S2andS3Figs). Because multigene families are known to evolve extremely rapidly in their genome struc-ture, obtaining the full genomes of species closer to humanP. vivax is fundamental for a better
understanding of its evolution, adaptation, and emergence in different host species. For Plas-modium parasites, most species-specific genes are part of large gene families, such as var genes
inP. falciparum or pir genes that are present in all Plasmodium genomes studied [19,20].
Table 2provides a summarized view of gene content and copy number of the main multigene families inP. vivax-like in comparison with P. vivax, P. knowlesi, and P. cynomolgi. Even if
cer-tain subtelomeric regions of our reference genomes (S2andS3Figs) are not complete, at least one copy of each major gene family was detected (Table 2). In comparison withP. vivax (and
expected because of the partial subtelomeric sequencing coverage), the number of copies in each family was generally lower or equal inP. vivax-like. For these families, all genes were
functional except for the Cytoadherence-linked asexual gene (clag) family. For the clag family,
all genes are functional except the one situated on chromosome 8 forP. vivax-like (confirmed
for both Pvl01 and Pvl06) (S4 Fig). Theclag family, strictly conserved in malaria parasites, is
an essential gene family in host–parasite interactions, playing a role in merozoite invasion, parasitophorous vacuole formation, and in the uptake of ions and nutrients from the host
Table 2. Number of detected copies of multigene family members in the genomes ofP. vivax-like (Pvl01 and Pvl06), P. vivax strains SalI and PvP01 [15],P. cyno-molgi B and M isolates [8,17], andP. knowlesi H strain [18].
P. vivax-like (Pvl01) P. vivax-like (Pvl06) P. vivax (PvP01) P. vivax (SalI) P. cynomolgi (B strain) P. cynomolgi (M strain) P. knowlesi (H strain) vir/pir (subtelomeric) 148 14 1,212 303 265 1,373 64 msp3 (central) 9 0 12 11 12 14 3 msp7 (central) 11 12 13 13 13 11 5 dbp (subtelomeric) 1 0 2 1 2 2 3 rbp (subtelomeric) 9 3 10 9 8 6 2 Pv-fam-a (trag) (subtelomeric) 34 36 40 34 36 39 29 Pv-fam-e (rad) (subtelomeric) 38 15 40 34 27 27 16
pst-A (subtelomeric and
central) 6 3 10 11 9 8 7 etramp (subtelomeric) 7 4 9 10 9 9 9 clag (RhopH-1) (subtelomeric) 3 2 3 3 2 2 2 PvSTP1 (subtelomeric) 4 0 10 9 3 51 0 Phist (Pf-fam-b) (subtelomeric) 20 12 84 64 48 54 15 sera (central) 13 7 13 13 13 13 7
ForP. vivax-like Pvl01 and Pvl06, a nonexhaustive list of family genes is represented because only partial genomes were obtained. Pseudogenized genes are included.
Abbreviations: CLAG, Cytoadherence-linked asexual gene; DBP, Duffy-binding protein;etramp, early transcribed membrane protein; msp, merozoite surface protein;
Phist,Plasmodium helical interspersed subtelomeric; pir, Plasmodium interspersed repeat; rbp, reticulocyte-binding protein; sera, serine-repeat antigen; STP1,
subtelomeric protein 1;trag, tryp-rich antigen.
plasma [21,22]. The pseudogenization of theclag gene on chromosome 8 for P. vivax-like
sug-gests that this species lost theclag gene during its adaptation to the ape host.
During the life cycle ofPlasmodium parasites, host RBC invasion is mediated by specific
interactions between parasite ligands and host erythrocyte receptors. Two major multigene families are involved in RBC invasion: the Duffy-binding proteins (dbp) and the
reticulocyte-binding proteins (rbp) multigene families [23]. DBP is a protein secreted by the micronemes of the merozoite stage that binds to the Duffy Antigen Receptor for Chemokine (DARC) to invade RBCs.P. vivax is characterized in its genome by 2 dbp genes (dbp1 on chromosome 6
anddbp2 on chromosome 1) that seem to be essential for RBC invasion, as demonstrated by
their inability to infect individuals not expressing the Duffy receptor on the surface of their RBCs (i.e., Duffy-negative individuals) [24–26]. In the reference genomes ofP. vivax-like
(Pvl01 and Pvl06), we observe thedbp1 gene as in the P. vivax genome PvP01 (Table 2) and also in the otherPlasmodium species (P. knowlesi and P. cynomolgi); however, we did not
observe thedbp2 gene (no read obtained mapping to this region). This observation was
con-firmed in the other genotypes sequenced in this study. Knowing that gorillas and chimpanzees are all described as Duffy positive [10], we propose thatP. vivax-like parasites infect only
Duffy-positive hosts, which could be associated with the absence of thedbp2 gene. This would
be in accordance with the fact that the only described transfer ofP. vivax-like to humans was
in a Caucasian Duffy-positive individual [9] and that no transfers ofP. vivax-like have been
recorded in Central African Duffy-negative populations despite the fact that they live in close proximity to infected ape populations [27].
rbp genes encode a merozoite surface protein family present across all Plasmodium species
and known to be involved in RBC invasion and host specificity [23]. Amongrbp family genes,
3 gene classes (rbp1, rbp2, and rbp3) exist and are associated with the ability of Plasmodium
parasites to invade different maturation stages of RBCs. In this study, comparison of the orga-nization and characteristics of therbp gene family between P. vivax, P. vivax-like, P. knowlesi,
andP. cynomolgi (Fig 1andTable 2) first reveals that gene classes RBP2 and RBP3 are ancestral to the divergence of all these species exceptP. knowlesi. Second, an expansion of the rbp2 class
is observed in theP. vivax/P. vivax-like/P. cynomolgi lineage (Fig 1A), suggesting that, in this lineage, specific expansion likely occurred during the evolution of these species. Finally,rbp3
genes, which are supposed to confer the ability to infect normocytes, are functional in all spe-cies except inP. vivax (for which the gene is pseudogenized in both SalI and PvP01 strains),
suggesting thatP. vivax lost the ability to infect normocytes or has developed an ability to
infect specifically only reticulocytes during its adaptation to human RBCs (Fig 1A and 1B).
Phylogenetic relationships to other Plasmodium species and
divergence time
Conservation of the gene content betweenP. vivax-like with the other primate-infective Plas-modium species has enabled us to reconstruct with confidence the relationships between the
different species and to estimate the relative age of the different speciation events. This analysis confirmed the position ofP. vivax-like as the closest sister lineage of P. vivax (Fig 2).
Regarding the estimation of divergence times using genomic information, different meth-ods were recently used forPlasmodium, such as the one implemented in Generalized
Phyloge-netic Coalescent Sampler (G-PhoCS) [28] or the one developed by Silva and colleagues [29]. G-PhoCS uses a Bayesian Markov Chain Monte Carlo (MCMC) approach to infer, based on the information provided by multiple loci, the divergence time between species. This method has been applied in 2 recent studies forPlasmodium parasites—one aiming at estimating the
malariae-like [14], the other to estimate the divergence time within theLaverania subgenus, a subgenus
includingP. falciparum and all its closest ape relatives [13]. The Silva method is based on the estimate of the sequence divergence in different proteins and comparison of this divergence measured between different lineages [29]. In this method, the regression slope of the diver-gence between the proteins in 2 lineages reflects their relative age. The advantage is that it does not rely on an estimate of mutation rate. Finally, it has already been used in a recent study esti-mating avian and primatePlasmodium species divergence times [30]. Here, without calibration points and a good estimation of theP. vivax and other Plasmodium species substitution rates,
Fig 1.rbp genes in P. vivax-like and P. vivax. (A) Maximum likelihood phylogenetic tree of all full-length rbp genes in P. vivax-like Pvl01 (in blue),P. vivax SalI and PvP01 strains (in green), P. cynomolgi B strain, and P. knowlesi H strain (in black). Bootstrap values, calculated by RAxML
bootstrapping posterior probability, are indicated. The different subclasses ofrbp are indicated as rbp1, rbp2, and rbp3. The black stars indicate
pseudogenes. The animal pictograms indicate the primate host. (B) Table representing the number of variants (including the ones that are pseudogenized) observed in eachrbp subclass in P. vivax-like (Pvl01), P. vivax (SalI and PvP01), P. cynomolgi (B and M strains), and P. knowlesi (H
strain). Pseudogenes detected among each subclass ofrbp are indicated within each subclass between brackets. The alignment of the rbp sequences
with their accession numbers indicated and the inferred RAxML tree are available as the supplemental files inS2 Data. RBP, reticulocyte-binding protein.
our aim was to evaluate the divergence time betweenP. vivax and P. vivax-like relative to the
divergence time between the other primate–nonprimatePlasmodium species pairs. To evaluate
the influence of the approach on the estimations, we used 2 strategies to estimate divergence time betweenP. vivax and P. vivax-like relative to the other divergence events within the
tree: the Silva method [29] and the RelTime method (seeMaterials and methodssection for a description of the two methods) [31,32]. We first used the Silva method [29], checking for the influence of the model of evolution and the reference species pair on the analysis (seeMaterials and methodssection). Because neither influenced the results of the relative ages of species pairs (seeFig 2,S5 FigandS4 Table), we only report the results for the Silva method using the JTT (Jones, Taylor, and Thornton) model of evolution and theP. ovale curtisi–P. ovale wallikeri
pair as the species pair of reference as in Bo¨hme and colleagues inFig 2[30]. While the 2 meth-ods are mostly in agreement (with RelTime showing larger confidence intervals [CIs]), they show a discrepancy concerning the relative divergence time of theP. malariae and P. malar-iae-like parasites (seeFig 2andS4 Table). However, both methods suggest that the evolution ofP. vivax in humans did not occur at the same time as the other human malaria agents of the Plasmodium genus (i.e., P. malariae and P. falciparum). The analyses show that the time of the
Fig 2. Relative divergence dating betweenP. vivax and P. vivax-like. Maximum likelihood phylogenetic tree of 13 Plasmodium species, including P. vivax and P.
vivax-like. The analysis was based on an alignment of 2,784 one-to-one orthologous groups across the 13 Plasmodium reference genomes (seeMaterials and methods section). Values indicated at the nodes are the 95% CIs of the relative splits estimated using the method developed by Silva and colleagues [29] (values in blue) and the RelTime method [69] (values in red;×100). Results are given for the analyses performed considering the JTT model of evolution. For the Silva method [29], we gave for the internal nodes the minimal and maximal values of the lower and upper limits of the 95% CI of all possible pairwise species combinations. The table gives the ratio between the relative divergence times of the human–nonhumanPlasmodium species pairs. The final alignment of the 2,784 one-to-one orthologues—excluding
ambiguities, lox-complexity regions, and poorly aligned regions—and the tree are available as supplemental files inS3 Data. CI, confidence interval; JTT, Jones, Taylor, and Thornton; Pf,P. falciparum; Pkn, P. knowlesi; Pm, P. malariae; Pml, P. malariae-like; Poc, P. ovale curtisi; Pow, P. ovale wallikeri; Pprf, P. praefalciparum; Pv, P. vivax; Pvl, P. vivax-like.
split betweenP. vivax and P. vivax-like happened before the divergence between P. falciparum
andP. praefalciparum (about 2.3 times earlier). It is unclear from our analyses whether the
divergence between the 2P. malariae species occurred before or after the divergence of the
other 2 human–nonhuman parasite pairs. Unlike us, a previous study estimated, using the G-PhoCS method, the relative dating of the split betweenP. reichenowi (a chimpanzee
para-site) andP. falciparum (P. praefalciparum was not available at that time) and that of the split
betweenP. malariae and P. malariae-like (note that similar estimates were recently obtained
using another set of data using the Silva method) [14] to have occurred at the same time. This discrepancy between methods suggests that the same strict molecular clock (which is a hypothesis of the Silva method) may not apply over the entire tree (especially for theLaverania
subgenus because of their extremely low GC content in comparison with otherPlasmodium
species). Whatever the method used, all these estimates nevertheless suggest that the evolution ofP. vivax in humans did not occur at the same time as the other human malaria species and
that the transfer ofPlasmodium parasites to humans may have happened several times
inde-pendently over the history of theHomo genus [33–37].
Relationships to worldwide human P. vivax isolates
To analyze the relationship between our 11P. vivax-like isolates and human P. vivax, we
com-pleted our dataset with 19 published humanP. vivax genomes (S1 Table) [38]. All sequencing reads were aligned against the PvP01 reference genome [15], and SNVs were called and filtered as described in the Materials and methods section. Maximum likelihood phylogenetic trees were then produced based on 100,616 SNVs. Our results clearly demonstrate the presence of 2 significantly distinct genetic clades (with a bootstrap value of 100) composed ofP. vivax-like
strains on one side and humanP. vivax isolates on the other side (Fig 3). This result differs from previous results suggesting that human strains formed a monophyletic clade within the radiation of apeP. vivax-like parasites [10].
One explanation for this difference with previous published results could be that it is due to a phenomenon called Incomplete Lineage Sorting (ILS) or to a lack of phylogenetic signal for phylogenies performed on a single or few genes. ILS is the discordance observed between some gene trees and the species or population tree due to the coalescence of gene copies in an ancestral species or population [39]. Such a phenomenon is often observed when species or population divergence is recent, which is the case forP. vivax/P. vivax-like [40,41]. ILS may thus result in the wrong conclusion ofP. vivax and P. vivax-like populations being intermixed
andP. vivax diversity being included in the diversity of P. vivax-like. A lack of phylogenetic
signal, which occurs frequently when species diverged recently, would have similar conse-quences. In order to test the implication of ILS or lack of phylogenetic signal, we generated a phylogenetic tree and a reticulated network on partial mitochondrial genomes ofP. vivax and P. vivax-like obtained in Liu and colleagues [10], Prugnolle and colleagues [9], and in the cur-rent study. These analyses show that partial mitochondrial genetic information is not enough to make a final conclusion on the origin ofP. vivax parasites (S6 Fig). When considering the polymorphism level of this portion of mitochondrial genomes, over the 2,483 bp, only 127 positions showed variability, 20 of them being a position specific to the outgroupP. cynomolgi.
Among the 107 variable positions found in theP. vivax and P. vivax-like samples, 87 were
sin-gletons, meaning that only 20 SNVs were shared by more than 2 individuals. This suggests that the discrepancy between our phylogeny and the one of Liu and colleagues [10] is probably because of a lack of phylogenetic signal. Indeed, in our study, the use of significantly more genetic information from throughout the genome, both in genic and intergenic regions, pro-vides a more accurate picture of the genetic relationships between the different parasite
species. Reducing our genetic data to single genes (as performed in previous studies) or a lim-ited number of SNVs also generates phylogenies in whichP. vivax is included within the
diver-sity ofP. vivax-like (seeS7 Fig).
Another explanation for the discrepancy between our results and the results from Liu and colleagues [10] could be that we are missing part of the diversity ofP. vivax-like given that
we obtained the genomes from a limited number of parasites isolated from a small popula-tion of apes in Gabon. Nevertheless, this hypothesis does not seem to hold in light of the tree and network produced using all the available mitochondrial sequences ofP. vivax-like and
our data, as isolates are distributed all over the currently known genetic diversity ofP.
vivax-like (S6 Fig).
Our results show thatP. vivax-like is composed of 2 distinct lineages: one including the 2
reference genomes (Pvl01 and Pvl06) and 7 other isolates that will hereafter be referred to as
Fig 3. ML phylogenetic tree with 1,000 bootstraps computed by alignment to theP. cynomolgi B strain genome,
based on 100,616 SNVs shared by 11P. vivax-like and 19 P. vivax samples. A position was considered an SNV if at least one sample carried a different nucleotide compared with the PvP01 reference. No missing data were allowed, and a minimum depth of 5 reads per position was considered. To overcome issues relating to multiple infections, we considered the dominant infection only by selecting the dominant allele (seeMaterials and methodsfor details). Bootstrap values superior to 70% are indicated. The host in which thePlasmodium parasite was detected is indicated by
the pictograms (human, chimpanzee, andAn. moucheti). This phylogeny showed the presence of a significantly
distinct clade (high bootstrap values associated with each clade) composed ofP. vivax-like strains on one side (light
blue) and humanP. vivax isolates on the other side (light green). Data (the alignment and the tree file obtained by the
ML phylogenetic analysis) can be found inS5 Data. ML, maximum likelihood; SNV, single nucleotide variant. https://doi.org/10.1371/journal.pbio.2006035.g003
P. vivax-like 1, and another one including 2 isolates (Pvl09 and Pvl10) that will hereafter be
referred to asP. vivax-like 2 (Fig 3). This sub-structuration ofP. vivax-like is confirmed in
the mitochondrial-reticulated network obtained using a larger number of isolates (seeS6 Fig). These 2 lineages may thus reflect an ancient split withinP. vivax-like or be the consequence
of a recent introgression or hybridization event betweenP. vivax-like and P. vivax in Africa.
A search of recent recombination events between lineages using SplitsTree (http://www. splitstree.org/) [77] does not support this latter hypothesis (S6 Fig).
Previous studies highlighted the high genetic diversity ofP. vivax-like populations in
com-parison withP. vivax worldwide [9,10]. In this genome-wide analysis of nucleotide diversityπ, we confirm thatP. vivax-like populations are significantly more diverse than P. vivax
popula-tions (P < 0.001; Wilcoxon test), with P. vivax-like samples showing nearly 10 times higher
nucleotide diversity (πP.vivax= 0.0012;πP.vivax-like= 0.0096). Such genetic diversity inP.
vivax-like strains in comparison with humanP. vivax has already been described in other studies
[9,10], suggesting thatP. vivax-like strains probably display extremely high genetic diversity
in Central African regions. This suggests thatP. vivax-like parasites of African great apes are
probably more ancient than the humanP. vivax strains and that the human P. vivax species (as
for humanPlasmodium species like P. falciparum) went through a bottleneck and only recently
underwent population expansion.
A still uncertain origin of P. vivax
Before the discovery of the apeP. vivax-like, the main hypothetical scenario concerning the
origin ofP. vivax was that of an “Out of Asia.” More specifically, it was considered that P. vivax
emerged in humans following its transfer to humans from Asian monkeys, as has been recently described forP. knowlesi [42]. However, the recent discovery ofP. vivax-like in African great
apes [10] and the analysis of their genetic characteristics led researchers to propose an “Out of Africa” origin ofP. vivax [10,36]. Based on phylogenetic analyses of partial mitochondrial genomes and nuclear sequences ofP. vivax and P. vivax-like parasites isolated from great apes
and humans, Liu and colleagues [10] suggested that all extant humanP. vivax parasites derived
from one single ancestor that was transferred from great apes to humans.
Our results do not bring any new evidence in favor of one or the other scenario. We think, nevertheless, that the origin ofP. vivax is more complex than recently proposed (a single host
switch from apes to humans in Africa) and that some previous data—especially those regard-ing the paraphyly ofP. like and the inclusion of P. vivax within the diversity of P.
vivax-like (the arguments used in support of an African scenario)—have been overinterpreted. The inclusion ofP. vivax into the P. vivax-like diversity based on only a couple of nuclear genes
and partial mitochondrial genomes is not a definitive proof of transfer ofP. vivax-like to
humans in Africa because alternative explanations can be provided. Indeed, as discussed above, such a phylogenetic pattern can be obtained because of ILS or because of a lack of phy-logenetic signal for the sequences used, 2 phenomena that are frequent when species diverged recently (which is the case forP. vivax and P. vivax-like). Our study indeed shows that
incon-gruent phylogenies may be obtained when one limits the analyses to a single or a couple of genes (see section above andS6andS7Figs).
In our mind, it is still impossible to conclude thatP. vivax has an African origin and that it
was transferred from African apes to humans, especially when several observations are more in favor of an Asian origin, such as the following: (i)P. vivax evolved in a clade of parasites
infecting Asian monkeys, and (ii) the highest genetic diversity ofP. vivax is observed in Asian
populations, and its diversity decreases toward Africa. This pattern of diversity is the opposite forP. falciparum, which has a well-established African origin [43]. For this parasite, the highest
genetic diversity is found in Africa and decreases toward Asia accompanying the human migration [43].
Because the data that are currently available are not easy to interpret and are somehow con-tradictory, we think that more complex scenarios regarding the origin ofP. vivax should be
envisaged in light of phylogenetic and population genetic evidences (as proposed in Prugnolle and colleagues [9]) and more genomic data need to be obtained in support of these scenarios.
P. vivax-specific adaptive evolution
Comparison of theP. vivax genome to its closest sister lineage (P. vivax-like) and to the other
primatePlasmodium parasites provides a unique opportunity to identify P. vivax–specific
adaptations to humans. We applied a branch-site test of positive selection to detect events of positive selection that exclusively occurred in theP. vivax lineage. Within the reference
genomeP. vivax-like (Pvl01), 418 genes exhibited significant signals of positive selection (S5 Table). In the humanP. vivax genome PvP01, the test allowed the identification of 255 genes
showing significant signals of positive selection (S6 Table). Among these genes presenting a significantdN/dSratio, 71 were shared betweenP. vivax and P. vivax-like (genes indicated in
orange in theS6 Table), including 56 encoding for proteins with unknown function and 15 encoding for proteins that are involved either in energy metabolism regulation (n = 9),
chro-matid segregation (n = 2), or cellular-based movement (n = 4).
We then took into consideration those 255 genes detected under positive selection inP. vivax
and compared them to those obtained inP. falciparum (172 genes under selection; see
Supple-mentary Table 4 in [13]). We identified a subset of 10 genes under positive selection in both the humanP. vivax and P. falciparum parasites (P < 0.05) (S7 Table). Among these 10 genes, 5 code for conservedPlasmodium proteins with unknown function and 3 for proteins involved in either
transcription or transduction. Interestingly, the 2 remaining genes under positive selection in these 2 humanPlasmodium parasites code for the oocyst capsule protein, which is essential for
malaria parasite survival in theAnopheles’ midgut, and for the rhoptry protein ROP14, involved
in protein maturation and the host cell invasion. These results suggest that these proteins could be essential for infection of humans or their vectors, and future studies should focus on the involvement of these proteins in human parasite transmission and infection.
Conclusion
Through technical accomplishments, we produced and assembled the firstP. vivax-like
refer-ence genomes, the closest sister clade to humanP. vivax—an indispensable step for a better
understanding of this enigmatic species. We established thatP. vivax-like parasites form a
genetically distinct clade fromP. vivax. Concerning the relative divergence dating, we
esti-mated that the divergence between both species occurred probably before the split betweenP. malariae species. This suggests that the transfer of Plasmodium parasites to humans happened
several times independently over the history of theHomo genus. Our genome-wide analyses
provided new insights into the adaptive evolution ofP. vivax. Indeed, we identified several key
genes that exhibit signatures of positive selection exclusively in the humanP. vivax parasites
and show that some gene families important for RBC invasion have undergone species-specific evolution in the human parasite, e.g.,rbp and dbp. Are these genes the keys to the emergence
ofP. vivax in the human populations? This pending question will need to be answered through
functional studies associated with deeper whole-genome analyses. Among the genes identified to be under positive selection, 2 were also identified to be under positive selection in the other main human malaria agent,P. falciparum, thus suggesting their key role in the evolution of the
study provides the foundation for further investigation intoP. vivax traits that are of public
health importance, such as features involved in host–parasite interactions, host specificity, and species-specific adaptations.
Materials and methods
P. vivax-like sample collection and preparation
P. vivax-like samples were identified by molecular diagnostic testing during a continuous
sur-vey of great apePlasmodium infections carried out in the Park of La Le´ke´di, in Gabon, by the
Centre International de Recherches Me´dicales de Franceville (CIRMF) [9]. In parallel, a survey ofAnopheles mosquitoes circulating in the same area was conducted in order to identify
poten-tial vectors of apePlasmodium [11]. Specifically, mosquitoes were trapped with CDC light traps in the forest of the Park of La Le´ke´di.Anopheles specimens were retrieved and identified using a
taxonomic key [44] before proceeding to dissection to isolate the abdomen. Samples were then stored at−20 ˚C until transportation to the CIRMF, Gabon, where they were stored at −80 ˚C until processed. Blood samples of great apes were treated using leukocyte depletion by CF11 cellulose column filtration [45].P. vivax-like samples were identified either by amplifying and
sequencing thePlasmodium Cytochrome b (Cytb) gene as described in Ollomo and colleagues
[46] or directly from samples already used for studies of otherPlasmodium species [13,14]. This allowed the detection of 11P. vivax-like samples, 10 from chimpanzees and 1 from an An. mou-cheti mosquito. Most of these samples were co-infected with other Plasmodium species and/or
probably with multipleP. vivax-like isolates (see below andS1 Table). The identification of intraspecificP. vivax-like co-infections was made by analyzing the distribution of the RAF [47].
Ethical approval
The animals’ well-being was guaranteed by the veterinarians of the “Parc of La Le´ke´di” and the CIRMF, who were responsible for the proceeding sanitary procedures (including blood collec-tion). All animal work was indeed conducted according to relevant national and international guidelines. These investigations were approved by the Government of the Republic of Gabon and by the Animal Life Administration of Libreville, Gabon (no. CITES 00956). It should be noted that our study did not involve randomization or blinding.
Genome sequencing
To overcome host DNA contamination, within 6 hours after blood collection in the La Le´ke´di park, each 5 mL chimpanzee blood sample was diluted with 1 volume of PBS and passed through CF11 cellulose powder columns to remove host leukocytes (i.e., leukocyte depletion) [45]. DNA was then extracted for the 11 samples using Qiagen Midi extraction kits (Qiagen) fol-lowing the manufacturer’s recommendations. Next, DNA samples were enriched using a WGA using REPLI-g Mini Kit (Qiagen) following the strategy described in Oyola and colleagues [48] to optimize WGA of AT-rich genomes from low DNA quantities. The Illumina isolates were sequenced using Illumina Standard libraries of 200- to 300-bp fragments, and amplification-free libraries of 400- to 600-bp fragments were prepared and sequenced on the Illumina HiSeq 2500 and the MiSeq version 2 according to the manufacturer’s standard protocol (S1 Table). Because of its high DNA content, the Pvl06 isolate was sequenced using PacBio chemistry. After a greater than 8-kb–size selection of DNA fragments using the BluePippin Size-Selection System (Sage Science), the library for this Pvl06 sample was sequenced using P6 polymerase and PacBio chemistry version C3/P5 in 20 SMRT cells. Raw sequence data are deposited in the European Nucleotide Archive. The accession numbers can be found inS1 Table.
Assembly of
P. vivax-like genomes
Host DNA decontamination. To overcome host DNA contamination for the chimpanzee
blood samples, contigs obtained were compared to chimpanzee orAn. gambiae genomes
(CHIMP2.1.4 genome, accession number: GCA_000001405; and AgamP3 genome, accession number: GCA_000005575) using Basic Local Alignment Tool (BLAST). When more than 50% of the contig hit the chimpanzee reference genome with more than 95% identity, the contig was removed from the analysis.
Determination of multiplePlasmodium species infections. To identify and quantify
Plasmodium multispecies infections, Illumina reads from each sample were mapped against
the reference genomes of thePlasmodium parasite species infecting primates P. vivax PvP01
[15],P. cynomolgi M version 2 [17],P. knowlesi H strain [18],P. reichenowi PrG01, P. billcol-linsi PbilcG01, and P. gaboni Pgab01 [13] using BWA version 0.7.12 [49] with default options.
P. vivax-like multiple infections. Following Chan and colleagues [47,50], we examined the RAF distributions to evaluate the possibility of co-infections with multiple strains ofP. vivax-like. A U-shape of the RAF distribution—meaning that almost all positions carry either
the reference or a single alternate allele (i.e., RAF of 100% or 0%)—would suggest a single infection. In contrast, if we observe both the reference and alternate alleles at some positions, this would suggest the presence of several strains in the sample. For eachP. vivax-like sample,
Illumina reads were mapped against the PvP01 [15] and other chimpanzee-infecting Plasmo-dium species P. gaboni, P. billcollinsi, and P. reichenowi reference genomes [13] using BWA version 0.7.12 [49] with default parameters. We only kept properly paired reads mapping to the PvP01 genome and removed PCR duplicates. SNVs were called independently for each sample using Samtools mpileup followed by bcftools version 0.1.19 [51,52] and were filtered to remove positions matching at least one of the following conditions: a quality phred score 30, a coverage <20, or aP value for strand bias, mapping quality, or tail distance biases 0.001.
For each sample and each of the filtered positions, we calculated the percentage of reads carry-ing the reference allele to draw the RAF distributions.
P. vivax-like genome assembly. Two P. vivax-like genomes (Pvl01 and Pvl06) were
assembled from a co-infection with aP. malariae-like and a P. reichenowi (PmlGA01 sample
in Rutledge and colleagues [2017] [14]) for Pvl01 and from a co-infection withP. gaboni for
Pvl06 (PGABG01 sample in Otto and colleagues [13]). Briefly, the genome assembly of the Illu-mina-sequenced sample Pvl01 was performed using MaSuRCA [53], and the assembled con-tigs belonging toP. vivax-like were extracted using a BLAST search against the P. vivax P01
reference genome [15]. The draft assembly was further improved by iterative uses of SSPACE [54], GapFiller [55], and IMAGE [56]. The 3,540 contigs resulting from these analyses were then ordered against the PvP01 genome and theP. gaboni and P. reichenowi reference genomes
[13] to separate possible co-infections with a parasite species of chimpanzees from the Lavera-nia subgenus using ABACAS2 [57]. The genome assembly was further improved and anno-tated using the Companion web server [16]. BLAST searches of the unassembled contigs against the 3 reference genomes were performed before running Companion to keep the con-tigs with the best BLAST hits against PvP01 only. The PacBio assembly of Pvl06 was performed using the Hierarchical Genome Assembly Process (HGAP) [58].
Variant calling for additional samples
Read mapping and alignment. Nine additionalP. vivax-like samples were sequenced for
population genomics and polymorphism analyses (seeS1 Table). The dataset was completed with 19 globally sampledP. vivax isolates [38] for human versus great ape parasite compari-sons, and the Asian parasiteP. cynomolgi strain B [8] was used as the root for phylogenetic
inferences (seeS1 Table). BWA [49] was found to be as specific as reads fromLaverania
sam-ples did not map onto the referenceP. vivax genome PvP01 (Otto, pers. com., results not
shown). The 11 newly generatedP. vivax-like samples, together with the already published 19 P. vivax samples and the reference strain P. cynomolgi [8] reads, were therefore mapped against the PvP01 reference genome using BWA version 0.7.12 [49] with default parameters. We then used Samtools version 0.1.19 to keep only the properly paired reads and to remove PCR dupli-cates [49]. Data quality was evaluated by the means of the number of reads properly paired, the mean depth per site, and the proportion of sites covered by at least 1, 5, 10, or 20 reads (seeS3 Table). Only positions with a quality phred score 30 and aP value for strand bias, mapping
quality, or tail distance bias >0.001 were considered.
SNV discovery. For population genomics and polymorphism analyses, SNVs were called
independently for all 11P. vivax-like and 19 P. vivax samples from the sequence alignment
files generated using BWA [49] to map the reads against the PvP01 genome (see above) [15]. For the SNV calling, we used Samtools mpileup version 0.1.19 [49] (parameters–q 20 -Q 20 -C 50) followed by bcftools to remove indels (call -c -V indels) [49]. SNVs were filtered using VCFTools [59] to keep variants that have been successfully genotyped in 100% of individuals, with a minimum depth of 5 reads per position. Allelic frequencies were also estimated for each variant at each SNV in order to select the dominant allele in multiple infections (—minDP 5 – max-missing 1—freq). Then, the files in VCF format were transformed into tab-delimited text format, using the “vcf-to-tab” Perl module, and finally transformed into fasta format using a homemade Perl script that could be obtained directly by contacting corresponding authors.
Gene family search
For theP. vivax-like Pvl01 and Pvl06, P. vivax PvP01 and SalI, P. cynomolgi B strain, and P. knowlesi H strain genomes, gene variants were detected and counted using Geneious
soft-ware [60]. This allowed the presence/absence of the variants of the different gene families acrossPlasmodium species to be evaluated. For each gene family, the number of variants
identified in the 2 reference genomes Pvl01 and Pvl06 was confirmed by manually checking the presence/absence of these in the otherP. vivax-like genotypes obtained using ACT and
bamview [61,62].
Orthologous group determination and alignment
Orthologous groups across (1)P. vivax PvP01 [15],P. vivax-like Pvl01, P. cynomolgi B strain
[8], andP. knowlesi H strain [18] reference genomes and (2) the 13Plasmodium reference
genomes used for the phylogeny (the here-generatedP. vivax-like Pvl01, P. vivax PvP01 [15],
P. cynomolgi M version 2 [17],P. coatneyi strain Hackeri [63],P. knowlesi H strain [18],P. fal-ciparum 3D7 [64],P. praefalciparum G01 [13],P. reichenowi PrCDC [65],P. gallinaceum 8A
[30],P. ovale wallikeri PowCR01, P. ovale curtisi PocCR01, P. malariae PmUG01, and P. malar-iae-like PmlGA01 [14]) were identified using OrthoMCL version 2.09 [66,67] after an all-against-all BLASTp (E-value threshold: 1× 10−6). From these, we extracted different sets of one-to-one orthologous genes for the subsequent analyses: a set of 4,056 genes that included the one-to-one orthologues among the 4 restricted species—P. vivax, P. vivax-like, P. cyno-molgi, and P. knowlesi (the Pv4sp set)—and a set of 2,943 among the 13 Plasmodium species
considered here for the interspecies phylogenetic analysis (the Pl13sp set). The first set of orthologous groups identified was used for the detection of selection (see below); the second one was used to build the phylogeny and for the dating analyses.
Amino acid sequences of the one-to-one orthologues were aligned using MUSCLE [68] or MAFFT [69], respectively, for the first and second set of orthologous groups. Prior to aligning
codon sequences, we removed the low-complexity regions identified on the nucleotide level using dustmasker [70] and then in amino acid sequences using segmasker [71] from ncbi-blast, using default parameters. After MUSCLE/MAFFT alignments, we finally excluded poorly aligned codon regions using Gblocks (parameters: -t = p -b5 = h -p = n -b4 = 2) [72]. After going through all these alignment cleaning steps, we ended with a low number of missing data and no gap in our dataset of one-to-one orthologues (31 missing data over the 2,784 align-ments of more than 50 amino acids).
Divergence dating
Most of the current methods available to estimate the timing of species splits make strong assumptions on the evolutionary models and often require an accurate mutation rate or cali-bration points, which are not always available. Here, we estimated the relative divergence times ofPlasmodium species to free from temporal calibration of phylogenies and used 2
methods that do not depend on the estimation of a mutation rate: a method based on pairwise amino acid sequence divergences and Total Least Squares (TLS) regressions but assuming a constant rate of evolution across thePlasmodium lineages—the so-called Silva method [29]— and the RelTime method developed in Tamura et al. [31,32], which does not assume any spe-cific model for the rate of evolution.
We first built the phylogenetic tree of the 13Plasmodium species considered here for the
dating analyses using the software RAxML version 8.2.8 [73]. The tree was constructed using the concatenation of the 2,784 orthologous groups that have a final alignment of more than 50 amino acids (among the Pl13sp set of 2,943 groups). RAxML was called using the following options in order to automatically determine the substitution model that best fits the data model and so that 100 bootstraps were performed to assess the tree robustness: “-m PROT-GAMMAAUTO -f a -# 100.” The best amino acid model of substitution identified by RAxML for the dataset (the JTT model with empirical base frequencies JTT + F) was used in the subse-quent analyses.
The idea behind the Silva method [29] for relative age estimation is that the divergence between amino acid sequences in independent lineages is correlated and that the divergence regression slopes of the proteins of a species pair of reference with the divergence of the pro-teins of other species pairs reflect the relative age of those splits (see Silva and colleagues [29] for a detailed description of the method). Following this method, we obtained, for each of the 2,784 orthologous groups of the Pl13sp set that have a final alignment of more than 50 amino acids and for each species pair, the amino acid sequence divergencedAAthrough a pairwise
comparison using PAML version 4.7 [74], with the option “cleandata” set to 1 to remove sites with missing data from the dataset. The sequence divergencesdAAwere estimated using 4
dif-ferent models of substitution (JTT, WAG, LG, and Dayhoff) to evaluate their influence on the estimates of the relative ages. An R script from the authors of this method [29] allowed the estimation ofα, the slopes of the TLS regressions of the divergence of the proteins between every possible species pairs and the reference species pair, with the 95% CI by bootstrapping (n = 10,000 bootstraps). The slope α is an estimator of the relative age of the 2 considered
spe-cies, relative to the reference species pair. To evaluate the influence of the choice of the refer-ence species pair on the results, we used different species pairs to consider multiple divergrefer-ence references: We considered the relative distance of the split of the speciation of (i)P. vivax and P. knowlesi (reference pair considered in the original paper), (ii) the 2 P. ovale species, and (iii)
theP. malariae-like and P. malariae species.
Because the model of Silva assumes a strict molecular clock [29]—which would not apply to allPlasmodium species, specifically P. falciparum because of its extreme GC content in
comparison with otherPlasmodium species—we used the clock calibration-free method
RelTime [31,32] implemented in MEGA7 [75] to estimate divergence times when evolutionary rates vary. RelTime estimates relative rates for each branch of the tree without requiring a dis-tribution of rate heterogeneity. The relative rates of 2 sister lineages are estimated assuming that the time to their most recent common ancestor is equal in the presence of contemporane-ous sampling, and the rate of their ancestral branch is estimated as the average of their branch-specific relative rates. This framework is applied to subtrees of 3 or 4 taxa following a bottom-up strategy, from tips to the root to estimate relative rates and times, relative rates being scaled by setting the lineage relative rate of the ingroop root node to 1 (see Tamura and colleagues [32]). RelTime analysis was performed on the concatenation of the alignments of the same 2,784 orthologous groups we used to build the phylogenetic tree, removing sites with missing data (Complete-Deletion option) and using the same substitution model (JTT + F). As required by the RelTime analysis, we specifiedP. gallinaceum 8A [30] as the outgroup.
Phylogenetic tree of
P. vivax and P. vivax-like strains
The phylogenetic relationships of the RBPs in theP. cynomolgi–P. knowlesi–P. vivax–P.
vivax-like group of species was reconstructed using a maximum vivax-likelihood analysis using RAxML [73] with 100 bootstrap replicates. The tree was visualized using Geneious software [60]. To investigate relationships between theP. vivax and P. vivax-like populations, we constructed a
maximum likelihood tree using the filtered variant call set of SNVs limited to the higher allelic frequency genotypes identified within each sample using RAxML and PhyML (using general time-reversible [GTR] models) [73,76]. Trees were visualized using Geneious software [60]. All approaches showed the same final phylogenetic tree described in the “Relationships to worldwide humanP. vivax isolates” section.
Analysis of phylogenetic discordance by means of reticulated networks
We looked for signatures of introgression and/or ILS by performing a phylogenetic network using SplitsTree 4.14.6 [77] based on the alignment of a portion of the mitochondrial genome containing thecox1 and cytb genes, as in Liu and colleagues [10]. We included all but 2 (KF618538 and KF618534) samples from Liu and colleagues [10], all from Prugnolle and colleagues [9], and 3 of our samples with a high-quality–sequenced mitochondrial genome (Pvl08, Pvl09, and Pvl10), and we addedP. cynomolgi as the outgroup (GenBank accession
number: AB434919.1). Gaps and monomorphic positions were excluded for the analysis, and the dataset resulted in an alignment of 85 sequences of 127 variable positions. We conducted a reticulate network using the RECOMB 2007 [78] and the NeighborNet distance transforma-tion methods [79]. A phylogenetic tree based on the complete alignment, without exclusion of sites (i.e., an alignment of 2,530 bp), was constructed using FastTree 2.1.5 (maximum likeli-hood) implemented in Geneious [60] for comparison with results obtained from Liu and col-leagues [10]. To explore the influence of the amount of phylogenetic signal on phylogenetic reconstruction, we also performed phylogenetic trees as described in the previous section using a subset of SNVs: Trees were reconstructed using 100, 200, 300, 400, 500, 600, 800, 1,000, or 5,000 SNVs.
Genome-wide nucleotide diversity
The nucleotide diversity (π) is the average number of nucleotide differences per site between 2 sequences. This parameter is interesting to estimate because it gives valuable information about variation in prevalence and demographic histories of the parasites. For theP. vivax and P. vivax-like populations, we calculated the genome-wide nucleotide diversity (π) from VCF
files using VCFTools [59]. The means of nucleotide diversities was compared betweenP. vivax
andP. vivax-like species based on the Wilcoxon-Mann-Whitney nonparametric test.
Detection of genes under selection
In order to identify genomic regions involved in parasite adaptation to the human host, i.e., regions under positive selection, we performed branch-site tests. To search for genes that have been subjected to positive selection in theP. vivax lineage alone, after the divergence from P. vivax-like, we used the updated branch-site test of positive selection [80] implemented in the package PAML version 4.4c [74]. This test detects sites that have undergone positive selection in a specific branch of the phylogenetic tree (foreground branch) using Likelihood Ratio Tests (LRTs) to compare models allowing positive selection or not. The selective pressure is defined as the ratio (ω) of nonsynonymous (i.e., dN, keeping the amino acid) to synonymous (i.e.,dS,
changing the amino acid) substitutions (dN/dS). Under neutrality, rates of synonymousdSand
nonsynonymousdNsubstitutions are equivalent, so theω ratio is expected to be equal to 1.
Under purifying selection, the rate of nonsynonymous substitutionsdNis expected to be lower
than the rate of synonymous substitutiondSbecause selection will prevent amino acid
modifi-cations: Theω ratio observed is then under 1. This process is acting in reducing the fixation of
deleterious mutation. Finally, under positive selection, amino acid replacement will be chosen, with thedNrate expected to be superior to thedSrate, leading to an observedω ratio superior
to 1. This indicates the fixation of advantageous mutations. All coding sequences in the core genome with a one-to-one orthologous relationship amongP. vivax Sal1, P. vivax-like Pvl01, P. knowlesi H strain, and P. cynomolgi B strain were used for this test (i.e., the 4,056 gene Pv4sp
set of orthologous genes). We obtaineddN/dSratio estimates per branch and gene forP. vivax
andP. vivax-like lineages alone using Codeml (PAML version 4.4c [74]) with a free-ratio model of evolution. Genes with a significant signal of positive selection inP. vivax only were
compared to the ones obtained inP. falciparum from Otto and colleagues [13] (S4 Tablein Otto and colleagues [13]), in order to identify, e.g., essential proteins for human or vector infection.
The data are deposited in the Dryad repository (doi:10.5061/dryad.32tm1k4) [81].
Supporting information
S1 Table. Overview of theP. vivax-like and P. vivax samples. Plasmodium species, sample
ID, accession number, geographic location, year of collection,Plasmodium co-infections, host
species infected by thisPlasmodium parasite, bibliographic reference, and usage are indicated.
For multispecies co-infections, a sample was considered infected with aPlasmodium species
when the percentage of reads mapping to the reference genome of that species was >3% (see
S2 TableandMaterials and methodssection). (XLSX)
S2 Table. Sequencing andPlasmodium multispecies co-infections information. Host
con-tamination was estimated by mapping the reads against the human genome from the Genome Reference Consortium GRCh37 (Genbank accession: GCA_000001405) for all samples but Pvl09, for which reads were mapped against the reference genome ofAn. gambiae str. PEST
(Genbank accession: GCA_000005575). Those reads mapping the human or the mosquito genomes were discarded.P. vivax-like samples were analyzed for multispecies infections by
mapping reads against the reference genome of chimpanzee-infectingPlasmodium parasites.
S3 Table. Overview of the SNV calling of theP. vivax-like samples.p1, p5, p10, and p20
refer to the percentage of positions covered by at least 1, 5, 10, or 20 reads, after filtering to select high-quality positions (quality phred score 30 and aP value for strand bias, mapping
quality, or tail distance bias >0.001; seeMaterials and methodssection for additional informa-tion). SNV, single nucleotide variant.
(XLSX)
S4 Table. Overview of the dating analyses under the JTT model of evolution. (A) Influence
of the choice of the species pair of reference for the divergence dating method of Silva and col-leagues [29]. The slopes of the regressionsα of the protein divergence between each species pair and the protein divergence between the reference species pairs (representing the relative age of the species pair of interest relative to that of the reference pair considered) and the 95% CI are given in columns C–O. The ratio ofα measured considering PocGH01-PowCR01 or PmUG01-PmlGA01 as the reference pair toα measured considering PKNH-PvP01 as the ref-erence pair is given for each species pair, as well as the mean over the species pairs, the SD, and the coefficient of variation (SD/mean), in columns Q and R, respectively. (B) Relative diver-gence dates estimated using RelTime [31,32]. PVL,P. vivax-like Pvl01; PvP01, P. vivax PvP01;
PcyM,P. cynomolgi M; PCOAT2, P. coatneyi strain Hackeri; PKNH, P. knowlesi strain H;
PowCR01,P. ovale wallikeri CR01; PocCR01, P. ovale curtisi CR01; PmUG01, P. malariae
UG01; PmlGA01,P. malariae-like GA01; PF3D7, P. falciparum 3D7; PPRFG01, P. praefalci-parum G01; PRCDC: P. reichenowi CDC; PGAL8A, P. gallinaceum 8A.
(XLSX)
S5 Table.P. vivax-like PVL01 genome-wide signatures of selection. The branch-site test of
positive selection for orthologous genes product, gene length,dN/dSratio for PVL01 as well as
theP value for this test are indicated. Only significant tests are reported. The test was
consid-ered significant when theP value was below 0.05. In orange are indicated genes with
signifi-cantdN/dSratios betweenP. vivax PvP01 and P. vivax-like PVL01. dN/dS= 999 when no
synonymous difference is observed in the alignment of the gene considered and thus the ratio is estimated to be infinite.
(XLSX)
S6 Table.P. vivax PvP01 genome-wide signatures of selection. The branch-site test of
posi-tive selection for orthologous genes product, gene length, anddN/dSratio for PvP01, as well as
theP value for this test, are indicated. Only significant tests are reported. The test was
consid-ered significant when theP value was below 0.05. In orange are indicated genes with
signifi-cantdN/dSratios betweenP. vivax PvP01 and P. vivax-like PVL01. dN/dS= 999 when no
synonymous difference is observed in the alignment of the gene considered and thus the ratio is estimated to be infinite.
(XLSX)
S7 Table.P. vivax and P. falciparum genome-wide signatures of selection. The list of the 10
orthologous genes showing a signature of selection in bothP. vivax PvP01 and P. falciparum
3D7 is indicated, based on the branch-site tests of positive selection. The branch-site tests of selection for PF3D7 genes were performed by Otto and colleagues [13] (see Suppl. Table 4 from Otto and colleagues [13]); PF3D7 orthologous genes to PvP01 genes were identified using the PlasmoDB database (http://plasmodb.org/plasmo/).dN/dS= 999 when no
synony-mous difference is observed in the alignment of the gene considered and thus the ratio is esti-mated to be infinite.
